Dr Rakesh Sengupta, Assistant Professor, Psychology, SIAS has published a research paper titled Long-Range Temporal Correlations in Enumeration: A Theoretical Link from On-Center Off-Surround Recurrent Dynamics to DFA of Reaction Times, in Proceedings of International Conference on Computing Systems and Intelligent Applications (ComSIA 2026), Lecture Notes in Networks and Systems (Vol 2058), Springer, Cham.
Research Summary
The Core Question
When looking at small numbers of objects (1–4 items), the human visual system perceives quantity instantaneously and accurately (“subitizing”), but for larger numbers, it transitions to slower, approximate “estimation”. Whether these represent two separate cognitive faculties or shifts within a single neural circuit has long been debated in cognitive science.
Theoretical Framework
Using a biologically inspired On-Centre Off-Surround (OCOS) recurrent neural network, the study proves that this behavioral divide reflects a dynamical phase transition. Subitizing operates in a highly stable, “sub-critical” regime where fluctuations decay instantly (memoryless white noise). In contrast, large-number estimation pushes the network toward a critical boundary, producing “critical slowing down” and long-range temporal correlations (pink noise).
Empirical Validation
By applying Detrended Fluctuation Analysis (DFA) to continuous enumeration reaction times, empirical human data matched theoretical predictions (subitizing DFA exponent ≈ 0.50 vs. estimation DFA exponent ≈ 0.67). Crucially, participants with stronger dynamical persistence in estimation experienced a higher reaction-time cost when switching back to subitizing (r≈ 0.41).
